Accelerated decline in apolipoprotein E-ϵ4 homozygotes with Alzheimer's disease

The apolipoprotein E-ϵ4 (APOE-ϵ4) allele has been identified as a powerful genetic risk factor for the development of AD.^1,2 The apolipoproteins are glycoproteins involved in lipid transport that play a role in brain development, synaptogenesis, and response to neuronal injury.^3-8 The ϵ class of apolipoproteins (APOE) has three primary variants, designated as ϵ2, ϵ3, and ϵ4. With population frequency estimates ranging from 0.63 to 0.85, ϵ3 is the most frequent allele, whereas ϵ4 and ϵ2 alleles are more rare; ϵ2 frequencies are estimated to range from 0.03 to 0.13 and ϵ4 frequencies from 0.07 to 0.29.³ ϵ4 Homozygotes, who by definition possess twoϵ4 alleles, are thought to be at particular risk for developing AD; the estimated percentage of ϵ4 homozygotes who will develop AD ranges from 50% to 90%.³

APOE functions in the CNS that may potentially affect the pathophysiology of AD include sequestration of the protein tau and related formation of paired helical filaments,⁹ and deposition of beta amyloid.^2,10,11 Relative to other patients with AD, ϵ4 homozygotes show earlier age at onset, increased amyloid burden, and decreased acetylcholine levels-findings that suggest potential differences in disease severity or rate of progression.^2,12-14 Surprisingly, studies have demonstrated no genotype differences in rate of decline.^10,12,15-17 As a consequence of these negative findings, it has been suggested that APOE status relates to the onset or age of expression of AD but does not play a mechanistic role in disease progression.^15-17 This conclusion has important implications for determining whether the APOE-ϵ4 allele primarily affects susceptibility to AD or whether it contributes directly to the pathophysiology of AD. These negative studies, however, may be compromised by methodologic biases that commonly plague longitudinal epidemiologic investigations of AD. For example, several studies had insufficient sample sizes to examine individually the rarer genotypes such as genotypes including an ϵ2 allele or ϵ4 homozygotes. In addition, other studies used insensitive measures of cognition (e.g., the Mini-Mental State Examination [MMSE]) that are highly susceptible to floor effects as dementia progresses.

This study examined the rate of decline in newly diagnosed AD patients with different APOE genotypes, using a comprehensive measurement of cognitive status (Dementia Rating Scale [DRS]).¹⁸ The study includes the largest sample from which such longitudinal cognitive data have been published. We hypothesized that patients who were ϵ4 homozygotes would show an accelerated rate of decline relative to patients with other genotypes.

Methods. Patients. Participants were patients with probable AD, newly diagnosed using National Institute of Neurological and Cognitive Disorders and Stroke-Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA) and Diagnostic and Statistical Manual of Mental Disorders-III-Revised (DSM-IIIR) criteria,¹⁹ who were enrolled from 1989 until 1997 in a community-based AD patient registry maintained by the University of Washington and Group Health Cooperative, a large stable health maintenance organization located in Seattle. All patients were newly diagnosed because entry into the AD patient registry occurred within 1 year of their first medical diagnosis of AD. Patients with other CNS conditions or major psychiatric disorders were excluded using criteria that have been previously described.²⁰ All subjects received APOE restriction genotyping, resulting in four genotype groups: ϵ2/3(i.e., subjects with one ϵ2 allele and one ϵ3 allele; n = 14); ϵ3/3 (n = 75); ϵ3/4 (n = 82); and ϵ4/4 (n = 30).

Cognitive measures. The DRS was administered at baseline and then at annual follow-up visits for a period of 1 to 6 years. Mean length of follow-up was 2.47 years (SD 1.24). No differences were noted among groups in length of follow-up (p = 0.27). Patients received DRS testing until their MMSE scores fell below 10. All subjects had baseline and at least one follow-up DRS total score. DRS total scores represent the sum of scores from five subscales: attention, memory, initiation/perseveration, conceptualization, and construction. DRS total scores have a potential range from 0 to 144, with lower scores reflecting greater impairment.

APOE genotyping. DNA was derived from buffy coat preparations by the salting out procedure of Miller et al.²¹ APOE genotypes were determined using the PCR conditions of Emi et al. and the HhaI restriction digest method of Hixson and Vernier.^22-23

Statistical analysis. As described by Growdon et al., all available DRS total scores for each subject were entered as dependent variables into a regression analysis.¹⁶ From this regression analysis, a slope was derived for each subject reflecting annual rate of decline for the DRS total score. Thus, larger slope values reflected greater rate of decline. Slopes were then entered into an ANCOVA, with genotype (ϵ2/3, ϵ3/3, ϵ3/4, ϵ4/4) as the independent variable and DRS total score and age at entry as covariates. All possible interaction effects involving these terms were also considered; only the DRS by genotype interaction was found to be statistically significant and was added to the model. Each observation was weighted by the number of and length between follow-up visits, and within subject mean DRS total score was used to control for DRS total level, as previously described.²⁴ If APOE contributes to the pathophysiology of AD rather than just being a susceptibility factor, it would affect disease progression preclinically, as reflected in younger age at diagnosis, and during follow-up as reflected by a faster rate of DRS decline. Accordingly, age at entry may be considered to be a correlate of rate of decline rather than a covariate.²⁵ We therefore fit the model two ways-with and without the age term. The latter model intended to fully describe the relation between APOE genotype and rate of decline under the assumption that age at entry is a result of APOE genotype and not a covariate to be controlled.²⁵ To describe the interaction effects, covariate-adjusted least-squares means were calculated and compared using multiple F tests.

Results. No differences were noted among genotype groups in education (p = 0.87); estimated duration of symptoms at entry, as reported by family members (p = 0.63); MMSE scores(p = 0.56); or DRS total scores (p = 0.51) at entry into the AD patient registry. The ϵ4/4 group was significantly younger than all other genotype groups (table).

Assuming that APOE contributes to the pathophysiology of AD and predicts age at entry, the full effect of APOE on rate of decline may be better described by the model fitted without the age term.²⁵ With this model, rate of decline as measured by DRS slope differed according to genotype, with the effect modified by DRS score [F(3, 196) = 3.61, p < 0.014]. At the mean DRS score observed in our sample (DRS = 105), the ϵ4/4 group had a significantly increased rate of decline(11.9 points per year) relative to the ϵ2/3 group (5.8 points per year; p < 0.003), with similar trends noted in comparison with the ϵ3/3 (9.3 points per year; p < 0.076) andϵ3/4 (9.6 points per year; p < 0.055) groups(figure). In addition, rate of decline for theϵ2/3 group was significantly slower than the ϵ3/3 and ϵ3/4 groups (p < 0.047 and 0.027). At a lower DRS score (DRS = 80), even larger differences were observed among genotypes; the ϵ4/4 group had a increased rate of decline (22.2 points per year) relative to theϵ2/3 (9.7 points per year; p < 0.0006); ϵ3/4 (15.8 points per year; p < 0.021); and ϵ3/3 (18.2 points per year; p < 0.174) groups. Again, rate of decline for theϵ2/3 group was significantly slower than the ϵ3/3 and ϵ3/4 groups (p < 0.006 and 0.038). These data are presented in the figure.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Predicted annual rate of decline for each genotype group at Dementia Rating Scale (DRS) scores of 80, 105, and 130. Rate of decline is expressed in DRS points per year; thus, larger values reflect greater rate of decline. At DRS scores of 105, the covariate-adjusted mean slope for the ϵ4/4 group is significantly greater than the slope for the ϵ2/3 group (p < 0.003), with a similar trend noted in comparisons with the ϵ3/3 (p < 0.076) and ϵ3/4 (p < 0.055) groups. At DRS scores of 80, even larger differences are observed among genotypes; the ϵ4/4 group again had an increased rate of decline relative to the ϵ2/3 (p < 0.0006), ϵ3/4 (p < 0.021), andϵ3/3 (p < 0.174) groups. The ϵ2/3 group had a significantly slower rate of decline than all other groups at DRS scores of 80 or 105.

When age was entered as a covariate in the model, a similar estimated model was obtained. The genotype by DRS total term remained a significant contributor to the model [F(3, 185) = 3.71, p < 0.013]. In addition, older age predicted a slower rate of decline (p < 0.0002). Adding age to the model modestly attenuated the genotype effect for the ϵ4/4 group, who were an average of 3 years younger than the other groups. At the mean DRS score of 105, there was little difference in predicted rate of decline between the ϵ4/4 group (10.8 points per year) compared with the ϵ3/4 (9.4 points per year; p = 0.279) or ϵ3/3 (9.2 points per year; p = 0.237) groups. A significant difference remained between the ϵ4/4 and ϵ2/3 (5.8 points per year) groups (p < 0.009). The ϵ2/3 group had a significantly slower rate of decline than the ϵ3/3 or ϵ3/4 groups(p < 0.043 and 0.028, respectively). At DRS scores of 80, however, a significant difference in predicted rate of decline was demonstrated between ϵ4/4 (21.6 points per year) and ϵ2/3 (9.4 points per year; p < 0.006) or ϵ3/4 (16.1 points per year; p < 0.040) groups, with a similar pattern noted for the ϵ3/3 group (17.4 points per year; p < 0.138). The ϵ2/3 group again had a significantly slower rate of decline than the ϵ3/3 or ϵ3/4 groups (p < 0.008 and 0.020, respectively).

Discussion. As predicted, patients with AD who were APOE-ϵ4 homozygotes had the fastest rate of cognitive decline on the DRS compared with patients with other genotypes. Conversely, the patients with ϵ2/3 genotypes had the slowest rate of decline, that was significantly slower than any other genotype group. No evidence of a gene dose effect was observed for patients with the ϵ3/4 genotype when they were compared with patients with the ϵ3/3 genotype; the rates of decline for these two groups were nearly identical.

Our results differ from several investigations of the effects of APOE-genotype status on disease progression, in which no differences in rate of progression were observed among genotype groups. For example, two investigations examined changes in cognitive tests such as the Cambridge Cognitive Examination and MMSE scores in small groups (e.g., n = 50 to 70) of patients with AD who were divided into ϵ4 carriers and noncarriers. They observed no differences between groups.^12,17 Growdon et al.¹⁶ followed up 66 patients with a battery of neuropsychologic tests and found no difference in rate of progression amongϵ3/3, ϵ3/4, or ϵ4/4 groups. Dal Forno et al.¹⁵ observed faster progression in an ϵ3/4 group when compared with a non-ϵ4 group, with no such differences observed for the ϵ4/4 group. Failure to observe changes in some studies may be due to the relative insensitivity of the measures in assessing change at lower levels of function. In addition, most previous studies used samples that 1) may be too small to adequately detect differences in group performance on these relatively insensitive tests, or 2) combined ϵ3/4 and ϵ4/4 subjects into a single group. The single investigation with a sample size sufficient to examine individual genotypes (n = 153) did not include anϵ2/3 group and used a scale with a limited range of scores (the Information-Memory-Concentration subset of the Blessed Dementia Scale).^10,26

The current study includes the largest sample of AD patients for whom longitudinal cognitive data have been reported. In addition, rate of decline was assessed with the DRS, a measurement tool that has a broad range of scores even at lower levels of function. The large sample size permitted separate analysis of the more rare ϵ2/3 and ϵ4/4 genotypes. The importance of individual genotype analysis becomes evident when one considers that the ϵ2/3 subjects showed the slowest rate of decline; thus, the distribution of scores in any study that did not include subjects with this genotype may be skewed in a negative direction, making it harder to detect poorer performance in ϵ4 homozygotes. Similarly, given that theϵ3/4 group showed significantly slower decline than the ϵ4/4 group, the distribution of scores from a group combining the two genotypes would be skewed in a positive direction because of the better performance of the ϵ3/4 group, making it difficult to detect a difference between the combined ϵ4 and non-ϵ4 groups.

Our results also differ from those of two investigations reporting slower rate of progression on the MMSE with increasing ϵ4 gene dose. In one study, baseline scores were estimated statistically, which may have affected the pattern of results.²⁷ The second study of 99 patients with AD reported slowed progression on a modified MMSE in patients with at least one APOE-ϵ4 allele when average conditional change scores were estimated using growth-curve analysis.²⁸ Subjects in this study, however, were drawn from a research-based population rather than the community-based population included in the present study.

A statistical issue that may also explain differences in results among studies is the use of age at onset as a covariate. If ϵ4/4 status is associated with accelerated progression of AD, when averaged these patients will show earlier onset of disease. If two individuals begin to deteriorate cognitively at the same age, the individual with the faster rate of progression will cross the threshold from preclinical to diagnosable disease at a younger age. The younger age at onset reported frequently for AD patients who are ϵ4 homozygotes in this and other studies may thus be causally related to their faster rate of progression. Using age as a covariate in this instance will remove some variance actually due to the effect of APOE genotype and produce a misleadingly weakened result.²⁵ Consensus has not yet been reached regarding this issue, but it is clear that careful consideration of the relation between age and genotype is needed to develop meaningful statistical models of disease progression.

Two potential confounds that may affect generalizing the current results include adequate length of follow-up and representativeness of the community-based sample. Regarding the former, the length of follow-up in our study ranged from 1 to 6 years, with a mean of 30 months (2.5 years). This length of follow-up is comparable to or exceeds that reported by other investigators, such as Gomez-Isla et al. (31.8 months) and Growdon et al.(22.8 months).^10,16 Similarly, the frequency of the individual APOE genotypes observed in this sample are comparable to other samples; in particular, we observed that the ϵ2/3 andϵ4/4 groups comprised 6.8% of our sample and 14.8% of our sample, respectively. Gomez-Isla et al. reported that 5% and 16% of their sample hadϵ2/3 and ϵ4/4 genotypes, respectively.¹⁰ Comparable percentages were also observed by Tsai et al. (6.5% forϵ2/3, 11.7% for ϵ4/4).²⁹ Thus, both length of follow-up and APOE genotype composition are similar for our study and other published reports, suggesting that our finding of accelerated cognitive decline for ϵ4/4 patients is not due to methodologic or sampling biases.

The question of whether APOE genotype affects the clinical course of AD is important for understanding the mechanistic role of APOE. APOE-ϵ4 is thought to be related to one of two pathologic mechanisms. First, theϵ4/4 genotype is associated with a greater density of senile plaques, presumably as a consequence of increased Aβ deposition.^2,10,30 Second, interactions between APOE-ϵ4 and tau are hypothesized to result in increased neurofibrillary tangle (NFT) formation.³¹ Given that NFT distribution and density appear more closely related to level of cognitive impairment in AD than senile plaque distribution, the present findings of increased rate of decline for ϵ4 homozygotes may support the role of APOE-ϵ4 in NFT formation.^31-34 An earlier age at onset for ϵ4 homozygotes was also observed, in concordance with many other investigators-a pattern that is thought to reflect the relation between APOE-ϵ4 and senile plaque formation.¹⁷

Although the finding of genotype differences in rate of decline suggests a mechanistic role for APOE, there is considerable overlap among theϵ4/4, ϵ3/4, and ϵ3/3 groups. This overlap indicates that APOE status may not be a clinically useful predictor of decline for patients with these genotypes. For patients with the ϵ2/3 genotype, the distribution of slopes is more clearly weighted toward slower rates of decline. Although further studies are needed to confirm this pattern, our results suggest that ϵ2/3 genotype is a potentially useful clinical predictor of slower rate of progression for patients with AD.

In summary, the current results suggest that APOE-ϵ4 homozygosity is associated with a faster rate of cognitive decline. The failure of previous studies to detect such a pattern may reflect insensitive measurement techniques and the inability to segregate individual genotype groups due to insufficient power. The findings of increased rate of cognitive decline for APOE-ϵ4 homozygotes and slower decline for AD patients with theϵ2/3 genotype provide further support for the hypothesis that APOE plays a mechanistic role in the neuropathologic progression of AD and is not simply related to disease onset.